Conclusions


The spectrographs worked relatively well during our first flight. Owing to the mount of the DroneSpex instrument not being strudy enough to prevent significant vibrations, data from the only BigSpex spectrograph were analyzed. The second flight was not successful because of turbulent weather, and the data from our first flight instead were selected for analysis. There should be some reconsideration in the use of flying as low as 900 meters owing to the high probability of boundary layer turbulence, in the vicinity of mountains, especially over Longyeardalen. A 900 meter, or lower, track should not be ruled out, but perhaps it would be better over ocean or along a wide valley.

The iridium experiment was also a success, but finding an alternative method of transferring files to the ground server would be advantageous, since FTP is taking some longer time is reasonable.

The goals for analyzing imaging spectrograph data were twofold: to construct an RGB composite image using three spectral samples, and to use a large number of samples in a supervised classification procedure. Data from the second 2000m path over Longyearbyen were used. RGB images made looked reasonably similar to the color photos taken, but periodic frame drops that occured while creating spectral images did result in occational color separation near boundaries with high contrast (i.e. building edges and shadows).

Owing to the large number of different objects in the scene, the classification results show that only the simplest object identifications could be made. For example, infrastructure could be separated from snow. But owing to the abundance of shadow in the scene, a number of subclasses had to be made that resulted in a number of overlapping and not numerically separate PDFs. The classification results would be greatly enhanced by a much more careful selection of the training set, including implementing a thresholding process for multiple spectral images in order to identify similar yet distinct classes, such as ice and snow.


(Finished 10.05.05 -MMK & JMH)